Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: KRRNKRR win in 290: a new record

Author: Joshua Shriver

Date: 07:56:15 10/20/05

Go up one level in this thread


Interesting, a thought on compression though. Why not use bz2? It's just as easy
to implement as zip and has better compression.

Josh

>
>I think, that my tbgenerator will be available soon. But it is useless without
>data mining programms.
>It is written in Assembler-586 (so the source code isn't interesting) and can be
>run only on Windows. The technical article should be much more interesting, than
>the source code.
>
>Now I have some statistics (for Dann Corbit's question):
>
>The total number of 3+4+5-man pawnless EGTs in yk-format = 130.
>The total size of 130 EGTs = 1,473,493,622 bytes.
>The total time of generation - less than 5 hours for my Celeron-1300
>computer.
>
>There are 100 3+4+5-man EGTs with pawns. For DTZ-metric compression is
>much better in comparison with DTM-metric.
>I already have written beta-version of pawn-generator. For knnkp the size of
>EGT is about 2 times less than in Nalimov's format.
>
>I expect, that the total size off all 230 EGTs will be about 4 GB.
>
>And 2 important features are:
>1. Now we are using zlib-library for compressing. Datacomp.exe gives better
>compression.
>2. For a data-mining program with 2-ply search only the smallest of wtm/btm
>EGTs is needed.
>
>Yakov



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.